A multisensoral approach for landuse classifications and grassland monitoring based on the possibility theory

Accurate landuse classifications derived from remote sensing data are of great importance as input data for hydrological models. However, during the vegetation period the problem often occurs that different landuses show similar spectral signatures. The authors report investigations carried out in the Bavarian alpine foreland, an area characterized by a complex landscape. In this area misclassifications mainly occurs between grassland and cereals. To solve this problems a method was developed to incorporate ancillary geo-information in a maximum-likelihood classification procedure using an approach based on the possibility theory